<![CDATA[Abstract: The Palestinian conflict has become a global issue that has triggered significant public responses, especially through social media platforms. This study aims to evaluate the performance of a hybrid Naïve Bayes–XGBoost model in classifying netizen sentiment toward the Free Palestine issue on platform X (formerly Twitter) in the year 2025. Data were collected using the X (Twitter) API with keywords such as #freepalestine and #savegaza, then processed through a series of preprocessing stages and sentiment labeling using a lexicon-based approach. The dataset was then split into 80% training data and 20% testing data to compare the performance of the baseline Naïve Bayes model and the hybrid model. The evaluation results show that the baseline Naïve Bayes model achieved an accuracy of 75.7% and an F1-score of 76%, while the hybrid Naïve Bayes–XGBoost model achieved a significantly higher accuracy of 95.5% and an F1-score of 96%. These findings indicate that integrating the two algorithms improves both accuracy and balance in sentiment classification, especially for unstructured and imbalanced data. This study recommends the use of hybrid models for public opinion analysis on social media and suggests further development using deep learning approaches. Keywords: Sentiment Analysis, Free Palestine, Naïve Bayes, XGBoost, Hybrid Model Abstrak: Konflik Palestina menjadi isu global yang memicu respons besar dari masyarakat dunia, terutama melalui media sosial. Penelitian ini bertujuan untuk mengevaluasi performa model hybrid Naïve Bayes–XGBoost dalam mengklasifikasikan sentimen netizen terhadap isu Free Palestine di platform X (Twitter) tahun 2025. Data dikumpulkan menggunakan X (Twitter) API dengan kata kunci #freepalestine dan #savegaza, lalu diproses melalui tahapan preprocessing, dan pelabelan menggunakan pendekatan lexicon-based. Selanjutnya, data dibagi menjadi data latih (80%) dan data uji (20%) untuk membandingkan performa antara model Naïve Bayes dasar dan model hybrid. Hasil evaluasi menunjukkan bahwa model Naïve Bayes dasar menghasilkan akurasi 75,7% dan F1-score 76%, sedangkan model hybrid Naïve Bayes–XGBoost mencapai akurasi 95,5% dan F1-score 96%. Temuan ini menunjukkan bahwa integrasi kedua algoritma mampu meningkatkan akurasi dan keseimbangan klasifikasi sentimen, khususnya pada data yang tidak terstruktur dan imbalanced. Penelitian ini merekomendasikan penggunaan model hybrid untuk analisis opini publik dimedia sosial, serta pengembangan lebih lanjut menggunakan pendekatan deep learning. Kata kunci: Sentimen, Free Palestine, Naïve Bayes, XGBoost, Hybrid Model]]>
Copyrights © 2025
